Diabetes-induced peroxynitrite impairs the balance of pro-nerve growth factor and nerve growth factor, and causes neurovascular injury
- First Online:
- Cite this article as:
- Ali, T.K., Al-Gayyar, M.M.H., Matragoon, S. et al. Diabetologia (2011) 54: 657. doi:10.1007/s00125-010-1935-1
- 949 Downloads
Diabetic retinopathy, the leading cause of blindness in working-age Americans, is characterised by reduced neurotrophic support and increased proinflammatory cytokines, resulting in neurotoxicity and vascular permeability. We sought to elucidate how oxidative stress impairs homeostasis of nerve growth factor (NGF) and its precursor, proform of NGF (proNGF), to cause neurovascular dysfunction in the eye of diabetic patients.
Levels of NGF and proNGF were examined in samples from human patients, from retinal Müller glial cell line culture cells and from streptozotocin-induced diabetic animals treated with and without atorvastatin (10 mg/kg daily, per os) or 5,10,15,20-tetrakis (4-sulfonatophenyl)porphyrinato iron (III) chloride (FeTPPs) (15 mg/kg daily, i.p.) for 4 weeks. Neuronal death and vascular permeability were assessed by TUNEL and extravasation of BSA-fluorescein.
Diabetes-induced peroxynitrite formation impaired production and activity of matrix metalloproteinase-7 (MMP-7), which cleaves proNGF extracellularly, leading to accumulation of proNGF and reducing NGF in samples from diabetic retinopathy patients and experimental models. Treatment of diabetic animals with atorvastatin exerted similar protective effects that blocked peroxynitrite using FeTPPs, restoring activity of MMP-7 and hence the balance between proNGF and NGF. These effects were associated with preservation of blood–retinal barrier integrity, preventing neuronal cell death and blocking activation of RhoA and p38 mitogen-activated protein kinase (p38MAPK) in experimental and human samples.
Oxidative stress plays an unrecognised role in causing accumulation of proNGF, which can activate a common pathway, RhoA/p38MAPK, to mediate neurovascular injury. Oral statin therapy shows promise for treatment of diabetic retinopathy.
5,10,15,20-Tetrakis (4-sulfonatophenyl)porphyrinato iron (III) chloride
Glial fibrillary acidic protein
Nerve growth factor
p38 mitogen-activated protein kinase
Proliferative diabetic retinopathy
Per os (by oral gavage)
Proform of NGF
Retinal Müller glial cell line culture cells
Tyrosine receptor kinase A
Vascular endothelial growth factor
Diabetes disturbs retinal homeostasis by activating glial cells, reducing neurotrophic support and increasing proinflammatory cytokines. These changes lead to accelerated cell death within the inner retinal and ganglion cells [1–3] and blood–retinal barrier (BRB) breakdown resulting in macular oedema and neovascularisation . Clinically, the disease manifests as diabetic retinopathy and eventually leads to impaired vision. The gold standard for treating diabetic retinopathy is limited to laser photocoagulation, an invasive procedure with serious side effects, as reviewed by Ali and El-Remessy . The lack of approved pharmacological treatment for diabetic retinopathy makes it essential to identify effective therapeutic approaches. Thus, understanding the molecular mechanisms that regulate retinal neurovascular dysfunction is of major clinical importance.
Traditionally, neurotrophins such as nerve growth factor (NGF) promote cell survival. However, we and others have reported paradoxical increases in levels of NGF despite neuronal death in clinical and experimental diabetic retinopathy [1, 6, 7]. Our recent study described a mechanism of peroxynitrite-induced impairment of NGF signal via tyrosine nitration of tyrosine receptor kinase A (TrkA), the NGF survival receptor, and upregulation of p75NTR (NTR is also known as NGFR), the neurotrophin receptor, causing retinal neurodegeneration in clinical and experimental diabetes . These findings prompted us to study the role of peroxynitrite in regulating release of NGF in diabetic humans and to elucidate how it induces neurodegeneration and BRB breakdown in experimental models. NGF is synthesised and secreted by glia as a precursor proform of NGF (proNGF), which is proteolytically cleaved, intracellularly by furin and extracellulary by the matrix metalloproteinase-7 (MMP-7), to generate the mature form (NGF) . Under oxidative stress and inflammatory conditions, the activity of proteases is altered, possibly resulting in accumulation of proNGF in injured neuronal and vascular tissue . However, little is known about the production pattern and activity of MMP-7, and the resulting imbalance between NGF and proNGF under diabetic conditions.
The molecular link between two well-documented early hallmarks of diabetic retinopathy, i.e. accelerated neuronal death and vascular permeability, remains obscure [1–3, 10–13]. While proinflammatory cytokines are known to activate Rho family kinases leading to actin cytoskeleton rearrangement and cell permeability in vasculature, Rho family members are also essential regulators of neuronal survival in the nervous system, as reviewed by Linseman and Loucks . Therefore, we examined the role of RhoA activation and its downstream target, p38 mitogen-activated protein kinase (p38MAPK), as well as the protective actions of atorvastatin, an inhibitor of the rate-limiting enzyme of cholesterol synthesis and activation of the Rho family GTPases. In the present study, we demonstrate how accumulation of proNGF and subsequent activation of RhoA and p38MAPK may play a central role in mediating diabetes-induced neurovascular injury. We also describe the molecular mechanisms by which blocking of diabetes-induced peroxynitrite formation in the diabetic retina exerts protective effects, such as (1) restoration of MMP-7 levels and activity, and thus of the balance between proNGF and NGF, and (2) preventing activation of RhoA and p38MAPK.
Human specimens were obtained with Institutional Review Board approval from the human assurance committee at the Medical College of Georgia and the Veteran Affairs Medical Center.
Clinical characteristics of participants providing aqueous humour samples
Diabetes duration (years)
Post mortem vitreous and retinal samples were obtained from the GA Eye Bank. Samples were identified as being from diabetic (more than 10 years of disease duration) or non-diabetic controls. Human vitreous samples were kept refrigerated and retinas were snap-frozen for further analyses. Demographics and retinal pathology of post mortem samples were recently published by our group .
Procedures with animals were performed in accordance with the Association for Research in Vision and Ophthalmology (ARVO) and the Veterans’ Affairs Medical Center Animal Care and Use Committee. Three sets (totalling 67 animals) of male Sprague–Dawley rats (∼250 g, 6 weeks old) from Harlan Laboratories (Indianapolis, IN, USA) were randomly assigned to control, treated control, diabetic or treated diabetic groups. Diabetes was induced by intravenous injection of streptozotocin (60 mg/kg). Detection of glucose in urine and blood glucose levels >13.9 mmol/l indicated diabetes. Treatment was initiated the day after confirmation of diabetic status and continued for 4 weeks for various endpoints. Groups treated with atorvastatin (Pfizer, NY, USA) received the drug by oral gavage (per os [p.o.]) at 10 mg/kg daily. Additional groups received (i.p.) the peroxynitrite decomposition catalyst 5,10,15, 20-tetrakis (4-sulfonatophenyl)porphyrinato iron (III) chloride (FeTPPs) (15 mg/kg daily or vehicle) and served as controls. Streptozotocin-injected animals had significant increases of blood glucose level (27.2 mmol/l) compared with controls (9.2 mmol/l). Treatment with atorvastatin had no effect on body weight or blood glucose levels in diabetic rats (27.3 mmol/l) or in treated controls (9.1 mmol/l). As shown before, treatment with FeTPPs had no effect on body weight or blood glucose levels in diabetic rats .
Transformed retinal Müller glial cell line culture cells (rMC-1) (V. Sarthy, Department of Ophthalmology, Northwestern University, Chicago, IL, USA) were previously characterised to express Müller cell markers . Cells were maintained in high glucose (25 mmol/l) or normal glucose (5 mmol/l) for 72 h or treated with 100 μmol/l peroxynitrite overnight (16 h). Medium for rMC-1 was DMEMF12 with 10% (vol./vol.) FBS and penicillin/streptomycin. Peroxynitrite stocks were prepared in 0.1 mmol/l NaOH (Millipore, Billerica, MA, USA). An equal amount of vehicle or decomposed peroxynitrite was used for control experiments.
Retinal protein extraction and western blot
Retinas were homogenised in radioimmunoprecipitation assay buffer to examine the abundance of various proteins as described by our group . Purchased antibodies included: proNGF and NGF (Alomone Labs, Jerusalem, Israel); p75NTR (Millipore); MMP-7 (AnaSpec, Fremont, CA, USA); and phospho-p38MAPK and p38MAPK (Cell Signaling, Danvers, MA, USA). Membranes were reprobed with β-actin (Sigma, St Louis, MO, USA) to confirm equal loading. Primary antibody was detected by horseradish peroxidase-conjugated sheep anti-rabbit antibody and enhanced-chemiluminescence (GE Healthcare, Piscataway, NJ, USA). The band intensity was quantified using densitometry software (Alpha Innotech, CA, USA) and expressed as relative optical density.
Immuno-colocalisation of proNGF in Müller cells
Fixed retinal optimal cutting temperature compound frozen sections were reacted with anti-proNGF antibody (Alomone Labs) and anti-cellular retinaldehyde-binding protein (CRALBP) for Müller cells (Affinity BioReagents, Rockford, IL, USA). Activation of Müller cells was assessed by immunostaining of glial fibrillary acidic protein (GFAP) (Affinity BioReagents). Images were collected using confocal microscopy (Zeiss, Jena, Germany).
Oxidative and nitrative markers
Slot-blot analysis was performed as described previously [1, 16]. Retinal homogenate was immobilised on to a nitrocellulose membrane that was reacted with antibodies against 4-hydroxy-2-nonenal (4-HNE) (Alpha Diagnostics, San Antonio, TX, USA) or nitrotyrosine (Calbiochem, San Diego, CA, USA), and optical density of samples was compared with controls.
Rho kinase activity
Rho kinase activity was assessed by pull-down assay. As previously described, retinas were homogenised in assay buffer . Bound Rho proteins were detected by western blot using anti-RhoA antibody (Millipore).
MMP-7 activity was determined in human aqueous humour samples or culture medum using fluorescent-labelled substrate (AnaSpec). Results were compared with a standard concentration of MMP-7 after activation of the enzyme by adding aminophenyl mercuric acetate (Sigma) for 1 h, followed by fluorescent substrate. Then MMP-7 activity was measured fluorimetrically with a plate reader (excitation 370 nm, emission 460 nm) (BioTek).
Blood–retinal barrier function
Integrity of the BRB was measured as previously described by our group [3, 12]. Plasma was assayed for fluoresciene concentration using a plate reader (excitation 370 nm, emission 460 nm) (BioTek). A standard curve was established using BSA-fluorescence in normal rat serum. Through serial sectioning (10 μm) and imaging (200 μm2) of retinal non-vascular areas, extravasation of BSA-fluorescence was detected.
Neuronal cell death
TUNEL assay was performed using immunoperoxidase staining (ApopTag-peroxidase) in whole-mounted retinas or ApopTaG in optimal cutting temperature (OCT) compound-frozen sections as described previously by our group . The total number of TUNEL horseradish peroxidase-positive cells was counted using light microscopy.
The results are expressed as means ± SEM. Differences among experimental groups were evaluated by ANOVA and the significance of differences between groups was assessed by the post hoc test (Fisher’s protected least significant difference) when indicated. Significance was defined as p < 0.05.
Diabetes causes imbalance of proNGF and NGF in experimental and human samples
Diabetes induces peroxynitrite formation in experimental and human samples
Diabetes stimulates proNGF accumulation in activated retinal Müller cells
Diabetes-induced peroxynitrite impairs MMP-7 activity in clinical and experimental diabetic retinopathy
Diabetes stimulates Rho kinase and p38MAPK activation in experimental and human samples
Atorvastatin prevents diabetes-induced retinal neurodegeneration
Diabetes-induced peroxynitrite formation causes BRB breakdown
We have previously shown that diabetes causes BRB breakdown in streptozotocin-induced animal models of diabetes [3, 12, 13]. However, the causal role of peroxynitrite has not been investigated. Hence, we performed quantitative analysis using serial image analysis of retinal fluorescence intensity that was normalised to plasma fluorescence intensity for each animal. The results showed ∼2.4-fold increase in fluorescence intensity in diabetic retinas compared with controls. Representative images showed diffuse and prominent fluorescence throughout the diabetic retinal parenchyma (Fig. 6c). Treatment with FeTPPs exerted similar vascular protective effects to those of atorvastatin and prevented diabetes-induced BRB breakdown.
A growing body of evidence supports the concept that diabetes disturbs homeostasis in the retina by activating glial cells, reducing neurotrophic and prosurvival inputs, and increasing proinflammatory cytokines, leading to accelerated cell death and vascular permeability, thus impairing vision, notions which have been reviewed previously [18–20]. In agreement, we and others have reported that increases in proinflammatory cytokines, including vascular endothelial growth factor (VEGF), NGF, TNF-α, intercellular adhesion molecule-1 and inducible nitric oxide synthase, in diabetic retinas were correlated with neuronal cell death and vascular permeability [1, 3, 12, 13, 21–23]. However, the role of neurotrophins in the diabetic eye remains unknown. Neurotrophins, including NGF, are secreted by glia as proforms (proNGF) that are proteolytically cleaved to mature ligands (NGF), exerting beneficial effects. Until recently, our knowledge of the release of neurotrophins in diabetic tissue had been limited to techniques such as ELISA assays and quantitative measurement of mRNA expression that could not differentiate proNGF from mature NGF. We and others have reported significant increases of NGF levels in diabetic rat retinal ganglia and serum/tears of patients with diabetic retinopathy [1, 6, 7]. Recently, the development and availability of specific antibodies for proNGF and mature NGF have facilitated a better understanding of neurotrophin release under pathological conditions. Analyses of NGF and proNGF from vitreous and aqueous humour samples from patients with >10 or >25 years of diabetes showed significant (three- or fivefold) accumulation of proNGF and 35% or 65% reduction of mature NGF compared with non-diabetic patients. Interestingly, patients with PDR underwent panretinal photocoagulation, which can induce release of proinflammatory cytokines such as IL-6, but not VEGF and SDF stromal cell-derived factor-1 (SDF-1) . However, the possible effects of panretinal photocoagulation on modulation of proNGF remain elusive. The imbalance between proNGF and NGF observed by us in diabetic human samples was also detected in rat retina lysate after 4 weeks of diabetes. In parallel, analysis of oxidative stress and peroxynitrite markers showed significant increases in 4-HNE adducts and nitrotyrosine formation in aqueous humour samples from diabetic patients and diabetic rat retinas compared with non-diabetic controls. These results lend further support to previous studies showing enhanced peroxynitrite formation in clinical and experimental diabetes [1, 3, 12, 22]. Treatment of our diabetic animals with atorvastatin exerted similar protective effects to FeTPPs in reducing peroxynitrite and 4-HNE adducts, as well restoring the balance between NGF and proNGF to normal levels. Accumulation of proNGF after injury has been detected in several diseases of the central nervous system such as Alzheimer’s, where pro-oxidative and pro-inflammatory milieus can reduce protease activity and NGF cleavage [8, 9]. Of note, we believe that this is the first report showing clinical and experimental evidence of significant accumulation of proNGF in the diabetic eye.
Our recent findings demonstrating the critical role of peroxynitrite in the paradox of accelerated neuronal and vascular death despite the significant increases in NGF and VEGF expression [1, 11, 25] prompted us to further characterise the role of peroxynitrite in gliosis and NGF release in the early stages of diabetic retinopathy. Peroxynitrite produced by glia cells is not toxic by itself, but causes activation and expression of proinflammatory cytokines. Supporting this, our previous studies have shown that Müller cells are not among the retinal cell population undergoing apoptosis at 4 weeks of diabetes . Instead, our current study demonstrated that Müller cells are activated, as evidenced by GFAP immunoreactivity and prominent colocalisation of proNGF at the end feet of Müller cells in diabetic rat retinas. The notion that Müller cells are activated, in response to diabetes-induced peroxynitrite, to secrete proNGF was further supported by in vitro results showing significant increases in proNGF levels in rMC-1 conditioned medium in response to high glucose or peroxynitrite, as well as inhibitory effects of FeTPPs or atorvastatin. Our results lend further support to previous studies demonstrating that peroxynitrite activated brain astrocytes to release proNGF [26, 27]. In addition to Müller cells, activated microglial cells can produce proNGF leading to neuronal death [28, 29]. Whether retinal microglial cells play a role in the initial proNGF secretion or sustain a regulatory loop of neurotrophin production during diabetic retinopathy remains to be explored.
While proNGF is cleaved intracellularly by furin, with mature NGF being trafficked to secretory vesicles , it is cleaved extracellulary by MMP-7 . Unlike most other family members of MMPs, MMP-7 is constitutively expressed by most adult tissue and its activity is redox-regulated . Hence it is conceivable to expect decreased protease activity under oxidative stress and inflammatory conditions. Our results showed diminished expression and activity of MMP-7 in diabetic retinas or in high glucose-treated Müller cells that were restored by FeTPPs or atorvastatin treatment. Supporting this, previous studies demonstrated diminished expression and activity of MMP-7 in streptozotocin-induced diabetic rats  and in vitro models in response to hyperglycaemia , or oxidative stress via oxidation of active sites of MMP-7 . Moreover, the activity of MMP-7 in aqueous humour samples from PDR patients showed a 50% reduction of MMP-7 activity. Interestingly, two human samples were excluded because of concurrent statin treatment and showed higher MMP-7 than samples from patients not on anti-hyperlipidaemia therapy. The role of peroxynitrite in modulating furin activity and hence intracellular cleavage of proNGF remains elusive and warrants future studies.
While mature NGF mediates neuronal cell survival through binding TrkA and p75NTR receptors, proNGF can promote neuronal apoptosis through binding p75NTR, as reviewed previously [34, 35]. In support of this, our previous work showed significant impairment of TrkA receptor function and upregulation of p75NTR levels in diabetic retinas from humans and rats , favouring activation of the latter in response to accumulated proNGF. It has been shown that overabundance of p75NTR constitutively activates RhoA, leading to neuronal death via activation of p38MAPK pathway in response to proNGF [28, 36–39]. In agreement with this, our results showed significant increases in active RhoA as well as p38MAPK in human and experimental diabetic retinas compared with non-diabetic controls (Fig. 5). Treatment of diabetic animals with FeTPPs or atorvastatin blunted these effects. RhoA is a major small GTP-binding protein that acts as a molecular switch to control a large variety of signal transduction pathways. In addition to its role in modulating neuronal survival, RhoA and its downstream target, p38MAPK, are involved in regulation of cell motility via reorganisation of the actin cytoskeleton and, as such, play a critical role in BRB breakdown [40–42]. Under diabetic conditions, BRB breakdown is thought to occur because of diabetes-induced oxidative and nitrative stress, resulting in increased activity of proinflammatory cytokines including VEGF and TNF-α, thus activating p38MAPK [1, 3, 12, 38, 43, 44]. Here, we show a potential role for proNGF as a new player that possibly causes BRB breakdown by directly activating the RhoA and p38MAPK pathway in vasculature or indirectly by causing neuronal death in the diabetic retina. Further studies involving the role of proNGF in BRB breakdown and elucidating the downstream signalling events are currently in progress by our group.
The current study investigated the neurological and vascular effects of two different drugs in diabetic animals: (1) the peroxynitrite decomposition catalyst FeTPPs, and (2) the lipid-lowering drug, atorvastatin. Although the results of the two drugs were in parallel, the neuro and vascular protective effects of atorvastatin were usually superior to those of FeTPPs, but did not reach significance, a fact possibly attributable to the pleiotropic effects of statins, including antioxidant effects and inhibition of GTP-binding proteins, in addition to their known cholesterol-lowering ability [13, 17, 45–48]. While previous studies examined the protective effects of statins in retinas from ischaemia/reperfusion models or in other tissue from diabetic animals, our study is the first to demonstrate the neuroprotective effects of atorvastatin in the diabetic retina. Although, the neuro and vascular protective effects of FeTPPs are significant, its therapeutic use is limited due to chronic administration of iron. On the other hand, the results of oral atorvastatin treatment were generated using a dose that produces peak plasma concentrations similar to those reported after 60–80 mg/day of atorvastatin in humans, and hence could be readily translated to patients with diabetic retinopathy .
We are indebted to C. Von Bartheld (Department of Physiology and Cell Biology, University of Nevada School of Medicine) for his careful review of the manuscript and helpful insights. This work was supported by the American Heart Association Scientist Development Grant (0530170N to A. B. El-Remessy), a research grant from Pfizer Pharmaceutical, a Juvenile Diabetes Research Foundation grant (2-2008-149 to A. B. El-Remessy) and the University of Georgia Research Foundation (to A. B. El-Remessy).
Duality of interest
The authors declare that there is no duality of interest associated with this manuscript.